Artificial intelligence with attention based BiLSTM for energy storage system in hybrid renewable energy sources. (August 2022)

• Record Type: Journal Article
• Authors: Faritha Banu, J.; Atul Mahajan, Rupali; Sakthi, U.; Kumar Nassa, Vinay; Lakshmi, D.; Nadanakumar, V.
• Is Part Of: Sustainable energy technologies and assessments. Volume 52:Part D(2022)
• Publication Date: 2022-08
• Subjects: Renewable energy sources -- Energy storage systems -- Deep learning -- Metaheuristics -- Attention mechanism -- Electricity price fluctuations; Renewable energy sources -- Periodicals; Energy development -- Technological innovations -- Periodicals; Electric power production -- Periodicals; Energy storage -- Periodicals; 333.79
• DOI: 10.1016/j.seta.2022.102334
• Languages: English
• ISSNs: 2213-1388

Abstract

Abstract: Hybrid renewable energy sources with energy storage systems (ESS) are becoming more prevalent in offering remote consumers with efficient, affordable, and dependable energy alternatives to traditional systems. The ESS is essential for the integration of hybrid RES in order to balance the supply of power with demand. Simultaneously, the price fluctuations and intermittent nature of RES led to the invention of ESS for RES. To do this, the recently established deep learning (DL) models can be utilized to optimize the ESS for RES for maximum performance. With this purpose in mind, this study presents an optimal attention-based bidirectional long and short-term memory (OABLSTM-ESS) technique for energy storage systems utilizing renewable energy sources. By resolving the optimization problem, the described OABLSTM-ESS technique attempts to minimize electricity costs while achieving productivity requirements. In addition, an attention based BiLSTM (ABLSTM) method has been created to take advantage of the inherent benefits of anticipated future processes. Moreover, the improved coyote optimization algorithm (ICOA) is used to best select the ABLSTM model's hyperparameters. The ICOA is created by merging the COA with the hill climbing (HC) idea, which helps to search for and improve the learning strategy while determining the candidate solutions for the COA. A vast array of simulation analyses are implemented to assure the improved performance of the OABLSTM-ESS approach. According to the results, the minimal CT for the OABLSTM-ESS approach was 0.53 min for the forecast of electricity prices, 0.50 min for the forecast of wind energy, and 0.47 min for the forecast of solar energy. The OABLSTM-ESS method outperformed the other methods, as evidenced by the data and discussion in their totality. Based on a comparison of the results, it was found that the OABLSTM-ESS method is better than recent approaches in many ways.